2006年-世界发展银行全球_Can_Migration_Reduce_Educational_Attainment__Evidence_from_Mexico_42页_375kb

Summary of "Can Migration Reduce Educational Attainment?" by David McKenzie and Hillel Rapoport

Core Content

This paper investigates the impact of migration on educational attainment in rural Mexico, focusing on the causal relationship between household migration and schooling outcomes. The study uses a demographic survey and an instrumental variable (IV) approach to account for potential endogeneity in migration decisions. It finds that migration has a significant negative effect on educational attainment, particularly for older children.

Main Findings

• Negative Impact on Schooling: Living in a migrant household is associated with a lower probability of completing high school, with a 13% reduction for males and a 14% reduction for females.
• Age-Specific Effects: The negative effect is most pronounced for 16 to 18 year-olds, with younger girls (12 to 15 years) also showing a significant negative impact.
• Migration Channels: The observed decrease in schooling is primarily due to:
  • Direct Migration: Young males in migrant households are more likely to migrate instead of attending school.
  • Increased Housework: Young females in migrant households are more likely to drop out of school to assist with household responsibilities.
• Remittances and Credit Constraints: For children with low-educated mothers, remittances may help relax credit constraints, slightly mitigating the negative effect of migration on schooling.
• Censoring and Nonlinearities: The study accounts for right-censoring in educational attainment data (children still in school) and nonlinearities in grade progression, using an IV-Censored Ordered Probit model.

Key Methodological Points

• Instrumental Variables: The authors use historical migration rates from 1924 as an instrument for current migration, to address endogeneity.
• Data Source: The study uses the 1997 Encuesta Nacional de la Dinamica Demografica (ENADID), which provides a more comprehensive measure of migration experience compared to the 2000 Mexican Census.
• Empirical Models:
  • Bivariate Probit Model: Reveals significant negative effects of migration on school attendance for 16 to 18 year-olds.
  • Two-Stage Least Squares (2SLS): Shows negative effects on educational attainment for 16 to 18 year-olds, but insignificant effects for younger children.
  • IV-Censored Ordered Probit: Accounts for censoring and nonlinearities, yielding more accurate estimates of the negative impact of migration on education.
• Controls: The authors include historical variables such as land ownership in 1910 and school infrastructure in 1930 to address potential confounding factors.

Theoretical Framework

• The paper presents a theoretical model where a household's optimal schooling decision is determined by the trade-off between the financial and non-financial costs of education and its returns.
• Migration can influence this decision in two main ways:
  • Remittances: Provide financial support, potentially increasing educational investment for the poor.
  • Incentive Effects: The prospect of future migration may reduce the incentive to invest in education, especially for older children who are more likely to migrate.
• The model suggests that the net effect of migration is likely negative for most rural Mexican children, even if remittances have a positive impact.

Policy Implications

• The study highlights the importance of considering non-remittance channels when analyzing the impact of migration on education.
• It underscores the need for policies that support educational investment in migrant households, especially for older children who are more vulnerable to the negative effects of migration.
• The findings suggest that migration networks may have long-term impacts on educational outcomes, even if the direct effect is small.

Conclusion

• The paper concludes that migration has a depressive effect on educational attainment in rural Mexico, with the strongest impact on older children.
• The use of historical migration data as an instrument allows for a more accurate estimation of the causal impact of migration, revealing that the effect is not merely due to remittances but also due to changes in child labor and schooling decisions.
• The results suggest that migration is not always beneficial for education and that its effects depend on the age of the child and the socioeconomic characteristics of the household.

Key Variables and Definitions

• Migrant Household: Defined as a household with at least one member aged 19 and over who has ever migrated to the U.S. for work or other reasons.
• Educational Attainment: Measured by grade-years of schooling completed, with seven ordered categories.
• Censoring: Accounts for children still in school, as their education is not yet completed.

Data and Sample

• The sample includes 20,388 children aged 12 to 18 from 12,980 households in rural Mexico.
• The study compares migrant and non-migrant households, controlling for community-level variables and historical data.

Limitations and Considerations

• The use of historical migration data as an instrument may not fully capture all aspects of migration's impact on education.
• The censoring issue is addressed using an ordered probit model, which better reflects the discrete and ordered nature of educational outcomes.
• The effect of migration is found to be more pronounced in poorer households, but still negative for most children.

Final Note

• The paper provides a nuanced understanding of how migration affects education, emphasizing the importance of accounting for multiple channels such as remittances, child labor, and changes in educational incentives.